-
Before Analysis: Explore Data
- Navigate to the C:/PollutionModelling directory in Windows and launch (double-click) Ozone.mxd. Within ArcMap, we will explore each layer's Attribute Table and Symbology.
Step #1: Geocode Home Address
- This step will plot a point at any address in Texas so we can analyze localozone concentrations after interpolating.
- Click (1) Geocode Address on the toolbar
- Select the Locations tab
- Type any address in Texas
- Click Find
- Right-click the top result, and select Add Point
Step #2: Enable Extensions
- This workshop requires both (1) Geostatistical Analyst and (2) Spatial Analyst. Ensure both are checked.
Step #3: Join Ozone Concentrations
- This step will Join a specific month's hourly ozone concentrations to the Air Monitoring Sites.
- Click (3) Join Ozone Concentrations.
- Under Ozone Table, use the drop-down menua nd select Jul$.
- We can use any month, but for the purposes of this first iteration, let's all use the same month.
- Under Output File, leave the default.
- If you have a red X, change the file name as this must not be the first time you are running this.
Step #4: Geostatistical Analyst
(The big step)
- This step has three parts: (1) Perform quick kriging using defaults, (2) Exploratory Spatial Data Analysis (ESDA), and (3) Rerun kriging removing trends.
Step #4A: Initial Kriging
- Click (4) Geostatistical Analyst/Geostatistical Wizard
- Kriging Step 1/5
- Under Methods select Kriging/Cokriging
- Under Input Data, and under Dataset:
- Source Dataset: ozoneMonitors
- Data Field: JUL__OZ_12
- Kriging Step 2/5
- Under Kriging Type, select Ordinary.
- Under Output Type, select Prediction.
- Kriging Step 3/5: Semivariogram/Covariance Modelling
- Semivariogram Values: Difference squared of the ozone measurements taken at pairs of sampling locations separated by different distances.
- This allows us to examine spatial relationships between measured points.
- Blue line is the model
- Red dots represent grouped pairs of points
- Three important terms:
- Nugget: Variability in the field data that cannot be explained by distance between the observations
- Sill: Maximum observed variability in the data
- Range: Point at which the semivariance stops increasing
- Ideally, we want a small nugget and a large sill.
- Kriging Step 4/5: Searching Neighborhood
- Click on the image to see which samplpe points are contributing to the predicted value.
- Expand Weights to see the legend.
- Next
- Kriging Step 5/5
- To judge if a model provides accurate predictions, verify that:
- The predictions are unbiased, indicated by a mean prediction error close to 0.
- The standard errors are accurate, indicated by a root-mean-square standardized prediction error close to 1.
- The predictions do not deviate much from the measured values, indicated by root-mean-square error that is as small as possible.
Step #4B: ESDA
- We will take a look at the (1) Histogram, (2) Semivariogram/Covariance Cloud, and the (3) Trend Analysis accessible via (4) Geostatistical Anslyst/Explore Data.
- Histogram
- Ensure Attribute is set to JUL__OZ_12.
- Click a bar to see the corresponding coordinates on the map.
- Definitely a southern pgrogression. I wonder why?
- Semivariogram/Covariance Cloud
- Ensure Attribute is set to JUL__OZ_12.
- X-axis shows the distance between pairs of points and Y-axis shows the variance
- We can see pairs of points with smaller distance have smaller variation, which is good.
- Trend Analysis
- Trend is a non-random component that can be represented by mathematics.
- Blue Line - North-South
- Red Line - East West
- * The trend lines are U-shaped, not linear. We need to ensure our kriging model follows this shape.
Step# 4C: Kriging Again
- Repeat Step 4A, this time selecting 2nd order polynomial as the Trend.
- Compare diagnistic statistics.
Step# 5: Buffer Home Address
- Ensure the Specified Distance and Units are set to 10 Miles.
- Click Next
- Hit the browse button, change the type to Shapefile
- Save
Step# 5: Buffer Home Address
- This step will calculate the average ozone concentration within the 10-mile radius around our address
- Input geostatistical layer: The kriging layer generated with the trend removal.
- Input raster or feature zone data: Your buffer
- Output table: Seelct a suuitable location/file name